成败型元件串联系统可靠性的Bootstrap置信下限某些问题的讨论

ON SOME ASPECTS ABOUT THE BOOTSTRAP LOWER CONFIDENCE BOUND FOR SERIES SYSTEM RELIABILITY COMPOSED OF COMPONENTS WITH PASS-FAIL DATA

Bootstrap方法是目前较为流行的统计方法之一。本文讨论成败型元件及成败型元件串联系统可靠性的Bootstrap置信下限的某些性质。本文讨论当元件试验数改变或成功数改变时,可靠性的Bootstrap置信下限的变化趋势,从而讨论此方法是否具有合理性。由本文讨论指出:对单个成败型元件,当试验数不变,成功数增大时,元件可靠性的Bootstrap置信下限有增大的趋势,对成败型元件串联系统,也有同样的结果,即Bootstrap置信下限在这方面具有直观合理性,但对单个成败型元件,当试验数增加K次,成功数也增加K次(即失败数不变)时,元件可靠性的Bootstrap置信下限有时反而会有下降的趋势,显然这是不合理的。而对试验数增加,成功数不变时,本文给出了可靠性置信下限变化趋势的关系式,并举例说明当试验数增加,成功数不变时,元件可靠性的Bootstrap置信下限有时会出现增大的趋势,这也是不合理的。

The Bootstrap method is a fairly recent development of statistics. In this paper, the tendency of variation of Bootstrap lower confidence bound of reliability during variation of number of tests or number of success both for single component or series system composed of components with pass-fail data are assessed. As for Bootstrap method, that there are intuitional reasonableness on some aspects and intuitional unreasonableness on some other aspects are presented.